Aa
Contrast
Text size

Blog / Data Work

Your Decade of Science Is a PDF Nobody Can Find

17 July 2026

Same river basin, same questions, five years on. The difference in citations was a public data portal.

The data exists and the science is solid, but nobody can find it, use it or cite it. Public dissemination is where a decade of careful monitoring reaches the world as static PDFs, months out of date, on a website that is disconnected from the database that produced them. Your findings do not get built on, your methods are not copied or improved and your data is not cited. That is not because it is not worth citing. It is because getting it means sending an email and waiting three weeks for a CSV with no metadata. Open source environmental data tools, such as the CKAN platform, are one of the practical ways out.

What it looks like

A water quality research institute had ten years of continuous monitoring across a major river basin: nutrients, dissolved oxygen, turbidity, biological oxygen demand and macroinvertebrate indices. It was one of the most complete freshwater datasets in its region, scientifically irreplaceable and publicly invisible.

Researchers who needed river basin data found the website, saw a summary report from eight months earlier and sent a data request. The institute received more than forty requests a month. Each one set off the same sequence: find the right subset, extract it, check station locations for anonymity, convert to the requested format, write documentation, send it and follow up when the recipient had questions. Two senior researchers spent a combined fifteen hours a week answering requests for data the institute wanted to share publicly anyway.

Meanwhile a university group studying the same river needed data for a grant deadline. They waited three weeks, got a CSV with column headers and no documentation of units or quality flags, spent another week trying to interpret it and finally gave up and designed their own monitoring campaign to collect what the institute already had. The institute's decade of data appeared in the literature far less often than that of a younger regional network with only two years of data but a proper open data portal. The difference was not the quality of the science. It was that one dataset could be found, downloaded, understood and cited without asking anyone for permission.

Four places where it goes wrong

Data is published as static PDFs. The usual public output is an annual report with indicators, charts, a methods section and tables in an appendix. It is published once, downloaded occasionally and rarely cited, and the data inside is locked in a format made for reading, not for analysis. A researcher has to find the PDF, locate the table, retype the values and hope the summary statistics are what they need. If they need another aggregation, period or area, they must ask you. That costs both sides time and often does not happen at all. A PDF displays data. It does not share it.

The website is cut off from the database. Most organisation websites were built by someone who has left, on a system that needs manual updates and shows an export from some time ago. The database holds current data and the website shows old data, and the gap grows every month. A policymaker looking for current indicators for a briefing finds numbers from eight months ago, and either uses outdated figures or goes elsewhere.

There is no way to cite the data. Scientific impact is measured in citations, and most monitoring programmes have no infrastructure for citing data: no DOIs, no versions that let a citation point at one release, no standard metadata that makes datasets discoverable through academic search, and no permanent landing page. Researchers who use your data cite the PDF report, not the dataset, so the dataset collects no citation record, and when a funder asks about data impact you can point to downloads of a report but not to direct reuse.

Every external request is handled by hand. Forty requests a month, each a manual job, is the most visible symptom and the most costly in staff time, but it is a symptom. The root problem is that there is no self service access, so everyone who wants your data has to ask a person. Much of what is extracted and emailed is data the institute wanted public. The manual process protects nothing. It consumes resources and creates friction that drives researchers to collect their own data instead of reusing yours.

What fixes it

Your data should be findable, downloadable, documented and citable by anyone, without asking for permission, at any hour and in any format.

Build an interactive public portal. A web interface lets people filter by place, period, parameter and quality level, preview the result and download it as CSV, JSON or Excel. The portal connects directly to your database, so it always shows current data and needs no manual extraction. CKAN, the open data platform used by governments around the world, or a custom interface on a standard web framework, can do this, and you build it once.

Update the website from the database. Instead of a static site updated by hand, the site queries the database directly. Current indicators refresh on their own, new data appears without anyone publishing it and the gap between database and website closes. This means separating the content layer of the site from its design layer, which is standard in modern web development but missing from many older organisational sites.

Give your data DOIs. Register datasets with a repository that assigns them, such as Zenodo, PANGAEA for oceanographic and environmental data, GBIF for biodiversity data or another repository suited to your field. Write metadata to international standards, create a permanent landing page for each release and version the data so citations can be specific. A dataset with a DOI is cited like a paper, so researchers cite the data and not only the report, and its impact becomes measurable.

Offer self service access with controls where they are needed. Most data that is requested manually should be public with no request process at all. Decide what truly needs controls, such as sensitive station locations, personal data or commercially sensitive information, and apply controls only there. For that data, replace the email process with a light registration and approval workflow: requesters register, accept terms of use and get access to the right tier automatically, and a person only reviews the sensitive cases.

The institute built a public portal with automatic updates and a full set of DOIs. Manual requests fell from more than forty a month to about three, all for sensitive subsets, and citations of its datasets rose clearly in the first year. A researcher who had earlier given up and collected their own data found the portal, downloaded what was needed in twenty minutes and published a paper citing the dataset, which started a new collaboration. The director said: "We spent ten years collecting data for the scientific community. This was the first year the scientific community could actually use it."

The pattern

Failures in public dissemination feel like a communications problem: a better website, a data manager with time to answer requests, a bigger budget to promote findings. They are an infrastructure problem that makes your science invisible to the people who could build on it, amplify it and fund it. Scientific impact is not only about what you discover. It is about whether anyone can find it, use it and cite it. A dataset that cannot be reached without an email and three weeks of waiting contributes less to science than one half its size that anyone can download in twenty minutes with full documentation.

Fix public dissemination and your science starts to compound in ways that internal improvements cannot. Researchers build on your data, papers cite your datasets, funders see impact you can demonstrate and collaborations start with people who found your portal. Your data deserves to be used. Make sure the infrastructure lets people use it.

Escaping Data Chaos

Also on LinkedIn

This article was first published in the Escaping Data Chaos series. The stories in the series are composites of real projects, anonymised.

Read it on LinkedIn

Recognise this problem?

One conversation to find out if we can help.

Book a free 15 minute call